For fixed cell labeling, I've used anti-giantin with great success. I haven't tried anti-TGN38. Adding to what Daniel commented, the dye-labeled BSA-complexed ceramides are a good option for live cell labeling, but it has to be trafficked to the Golgi with a carefully optimized protocol. A better option is the CellLight Golgi products (as long as the cells are mammalian and non-blood derived, this product will transiently transfect the Golgi with GFP or RFP and is fixable). There is some precedent for using the ceramides on fixed cells, but in my experience it can be tricky for selectivity from cell type to cell type.
I recommend anti-Giantin. In over expressed cells, Giantin labels better the Golgi area than TGN38. TGN38 labels Golgi derived vesicles as well, so it is not so specific...
I've used TGN38 with great success (trans-golgi; Sigma rabbit pAb) and GM130 (cis/medial-golgi; Millipore mouse mAb) in fixed cells. I've also used the CellLight Golgi GFP reagent for live cell imaging and found it to work really nicely, so long as your cells transduce well with Bacmam viruses.
I have very good experience with Cytopainter from Abcam. You can used on living cells and then fixate them afterwards. The procedure is straight forward and worked in our hands in different cell types equally well. It comes in different fluorescent colors as well, so you can combine it with fluorescently labeled proteins and antibodies. There are also markers for other cellular compartments, not only Golgi.
P. S. I am not working for Abcam, just very satisfied with the Cytopainter.
I ended up purchasing anti-giantin, so we will see how it goes. I went with an antibody just in case I need to do some co-IPs, but Cytopainter looks amazing if I want to do some time-lapse imaging down the road.